Notes Re: Illumina iDEA Challenge

From SnOwy - Ed's Wiki Notebook

Contents 1 iDEA Challenge 2 Presentation Notes 2.1 RNA-Seq data 2.2 Standard mRNA 2.3 Directional mRNA data 2.3.1 What this data looks like 2.3.2 looking for chimeric transcripts 2.3.3 next example: alternative splicing 2.3.4 data also can be used to show SNPs 2.3.5 Not really looked at yet 2.3.6 From the floor 2.4 DNA Methylation 2.4.1 Twin Study - multiple sclerosis present/absent 2.4.2 RRBS sample prep 2.4.3 Data analysis notes 2.4.4 CpG Island -- RRBS (mRNA) vs WGBS (whole genome) 2.5 Copy Number Variation 2.6 Small RNA 2.6.1 Small RNA Informatics 2.7 NGS (Next Generation System) Data Set -- functional genomics in Breast Cancer system 2.8 Getting the data / The challenge 3 Properties

iDEA Challenge

• Main: http://www.illumina.com/landing/idea/
• Judges: http://www.illumina.com/landing/idea/judges.ilmn
• Presentation (Webinar): https://illuminaevents.webex.com/ec0605lb/eventcenter/recording/recordAction.do;jsessionid=WMRyMJ0G1Jh0HnbPc2Zj5rZPQKcmvs1NHZ789l8P5pJRMs8wpy7v!-1892578192?theAction=poprecord&actname=%2Feventcenter%2Fframe%2Fg.do&actappname=ec0605lb&renewticket=0&renewticket=0&apiname=lsr.php&entappname=url0107lb&needFilter=false&&isurlact=true&rID=2513212&entactname=%2FnbrRecordingURL.do&rKey=5690ac10be4cf7bd&recordID=2513212&siteurl=illuminaevents&rnd=6198057508&SP=EC&AT=pb&format=short
• WebEx Player (Cisco): http://www.webex.com/downloadplayer.html

Presentation Notes

• dataset -- eight cancer cell lines
  • highly studied model systems
• assays were run on these eight cell lines -- breast cancer lines
  1. "RNA-Seq"
     • 50bp paired-end standard
     • 100bp single-read directional
  2. DNA Methylation
  3. CNV with low pass genomic DNA sequencing
  4. small RNA analysis

RNA-Seq data

• information rich
• gene count, expression profile
• SNPs, mutations
• chimeric transcript discovery
• gene annotation, discovery
• two methods used ...
  1. Standard mRNA-Seq (1.5 to 2 years old)
     • standard method
     • purify poly-A mRNA, randomly fragment
     • mRNA → cDNA → dscDNA
     • ligate sequencing adapters
  2. Direction mRNA-Seq aspects of mRNA-Seq + small-RNA kit
     • CIP = alkaline phosphatase
     • PNK = polynucleotide phosphate kinase
     • ligate 3'-, 5'- small RNA adapters
     • → cDNA
     • shotgun prep
• Directional: all of the reads are from a single strand
• Standard: reads from both strands
• both protocols: often do get many reads from the introns (normal) -- poly-A mRNA will occasionally retain these

Standard mRNA

• option to do paired-end sequencing: sequencing both ends of the mRNA
• dataset of paired fragment reads: ~90% mappable data
• variable insert sizes -- 100 to 120 bps, 120 bp on average; MB-468 has a lot of short-read contaminants

Directional mRNA data

• dataset of directional fragment reads: 100bp ~25%; 75bp ~60%
• less than half the reads are actually 100 bp length reads -- often 75 bp
• adapters -- influenced alignments -- used TopHat to create alignments
• adapters are a sequence that are found at the 3′ end of the read.

What this data looks like

• gene expression different across the eight samples
• ER+/ER- -like plot for expressed species between two cell lines shown (BT-20 vs MCF-7) -- scatter plot
  • ER is estrogen receptor -- refers to the presence of expression
  • gene shown is actually NRXN3
• sixty fold increase in reads -- (I think BT-20 has 948, MCF-70 has 15) -- this is a paired-end read
• hierarchical clustering of mRNA-Seq -- Gene Count Data
  • data shown for 18k genes

looking for chimeric transcripts

• Title: Chimeric transcript discovery by paired-end transcriptome sequencing
• paired-end data -- chimeric search -- genes have recombined to produce chimeric proteins on their own
• fusion transcripts
• MCF-7 fusion proteins: BCAS4 and BCAS3 (breast cancer amplified sequences 3 and 4)
• velvet algorithm: create contigs that contain fusion junctions
• search through these for contigs for fusion products that are real and that aren't real
• used blast with the fusion contigs to find such products
• result: discovered BCAS4_BCAS3_368_bp expressed product
• highly expressed contig with paired-end reads
• now considered a positive control for finding fusion products

next example: alternative splicing

• Title: Alternative isoform regulation in human tissue transcriptomes
• analyzed splices, 95% of multi-exon human genes show alternative splicing behaviour
• provided alignments using tophat / other data sets

data also can be used to show SNPs

• 35% of all SNPs overlap RefSeq genes have >10X coverage
• 10x coverage: up to 30 to 40 million 50 bp reads (yes! logarithmic) => roughly ten times the coverage of the genome.
• distinct SNPs -- most are non-coding (227k)
• 20k are non-synonymous coding (55%)!
• of the 20k, SIFT classification -- high-confidence damaging -- 24%, low-confidence damaging -- 25%, stop codon -- 2.4%
• many are novel discovered SNPs
• 47% are tolerated mutations; remaining are not scored.

Not really looked at yet

• Allele-specific expression
• Alternative splicing
• Gene discovery
• more?

From the floor

• there are 30 groups viewing the webinar

DNA Methylation

• Title: Genome-scale DNA methylation maps of pluripotent and differentiated cells
• epigenetic -- so called "fifth base" -- 5-methyl-C
• mostly CpG sites
• uncommon, clusters near 50% of gene promoters -- CpG islands
• cancer causes odd patterns of this
• Reduced Representation Bisulfite Sequencing (RRBS)
• genomic DNA restriction digested
  • size selection (150-175)bp, (175-225)bp -- insert size -- actual fragment sizes are 30-75, 75-125
  • bisulfite treatment used to alter 5-methyl-C bases from C → U => T
  • sequence before and after, the changes indicate where the bases have changed

Twin Study - multiple sclerosis present/absent

• Title: Genome, epigenome and RNA sequences of monozygotic twins discordant for multiple sclerosis
• highly correlated for methylation?!
  • wait doesn't that mean that we've supported the null hypothesis?
• okay, understand now -- it's the weird outliers that aren't on the main axis

RRBS sample prep

• 2.3M restriction sites in whole genome
• we only care about 0.75M of these fragments after size selection
• MSP1? MFP1? -- enzyme good for cutting CpG-islands
• 1.2% of the genome
• with the size selection, we get 91.4% of the genome we wanted
  • then where's the other 8.6%?
• pipeline:
• 50 bp read - may be shorter for shorter fragments
• C → T and G → A for fwd, rev strands due to base substitution for methylation detection
• aligned using 3 kinds of nucleotides: A, G, T.
• methylation rate given by proportion of C → T conversion in the before and after
• for well-methylated sites, we see 99.5% conversion
• typical methylation distribution: 10% low methylation, 20% high methylation -- remaining distribution demonstrates low probability
  • think: bimodal -- likely: due to cellular machinery errors: erroneous methylation of non-meth sites / non-meth of methylation sites.
• Note: CpG => C next to a G -- 5′-Cytosine-Phosphate-Guanine-3′ -- direct adjacency
• different cell lines have different methylation patterns
• experiment: correlation with mRNA-Seq patterns!
  • occasional correspondence with textbook: methylation α^-1 transcription

Data analysis notes

• two outputs
• one for individual sites
• one for each island
• island ID, number of CpG sites in an island
• number of sites that are covered
• depth of read covereage (average)
• level of methylation
• average of all methylation percentage

CpG Island -- RRBS (mRNA) vs WGBS (whole genome)

• lung cancer
• cAMP-mediated signaling -- cell death -- more at risk for cancer with increased methylation
• G-protein coupled receptor signalling -- increased methylation in cancer
• Axonal guidance signalling -- unrelated? pathway - increased methylation for cancer

Copy Number Variation

• low pass genomic DNA sequencing
• coverage of human genome is roughly 0.40x (40%)
• we're actually looking at specific regions of a chromosome
• allows us to describe using the proportion of hits, the likelihood of a particular copy number
• MCF7 - chromosome 17 - BCAS3 extra high copy number (30? fold, at least 25 fold)
• BT474 - chromosome 17 - ERBB2 extra high copy number (25 fold)
• using a sliding window, we estimate the copy number for each segment of a chromosome
• non-covered regions of the genome are likely centromeric regions -- looks like no coverage -- highly unique

Small RNA

• aka microRNA, miRNA
• pri-miRNA (transcription), pre-miRNA -- 60 - 70 nucleotides
• miRNA (mature) -- 21-25 nucleotides
• derived with cleavage of pre-miRNA

Small RNA Informatics

• piRNA: piwi-interacting RNA -- 26, 31 nucleotides
• siRNA: short interfering
• snoRNA: small nucleolar -- modifies rRNA, tRNA, snRNA
• snRNA: small nuclear RNA -- splices introns from mRNA
• 35 bp reads -- typical small RNA is 22 bp long
• adaptor trimming
• alignment for multiple-lengths
• multiple hits -- small sequences align different places in a genome
• lookup in library of alignments as well as raw genome
• sequence errors common as sequencing continues (into the adapter sequence?)
• library: lots of short sequences
  • miRBase (collection of miRNA)
• Flicker? Protocol / Software? What's Flicker?!
• Flicker outputs -- alignments to different databases including gDNA (genomic DNA), miRBase, both.
• miRNA hairpin
• Iso-miRs -- enzyme clipping miRNAs might be off by one or two nucleotides
• Cell lines -- MDA-MB-486 vs MDA-MB-231 small RNAs occur more often in the former?
• levels of hsa-miR-205 in the eight cell lines
• data counts the number of reads in each cell line and also in the normal genome

NGS (Next Generation System) Data Set -- functional genomics in Breast Cancer system

1. 2*50bp paired end mRNA-seq
2. 100bp stranded mRNA-seq
3. small RNA
4. DNA methtylation sequencing
5. CNV-seq analysis

• what can you do with this data in this competition?
• look for fusions, chimeras
• expression of mRNA vs miRNA
• methylation
• rna processing, alternative splicing
• patterns of mutations in cell lines
• iso-miRNA patterns in miRNA between cell lines
• mRNA v miRNA expression
• correlation of expression between data → what we're looking for in this challenge
  • mRNA, miRNA, methylation, CNV

Getting the data / The challenge

• sign up for the challenge
• will provide the HDD with data
• microsite just for the people in the competition
• does not include the human lung cancer methylation assay
• publication guidelines -- no guidelines yet, focus on competition; however, eventually will encourage publication later
• normalization: none done
  • for the methylation -- only regions which have 10X reads pass the threshold to be included as a percentage methylation
  • no other explicit normalization/thresholding ideas
• gschroth @ illumina . com

Properties

• Cytogenetics & CNV: http://www.illumina.com/applications.ilmn#cytogenetics
• Genome & CNV: http://www.illumina.com/applications.ilmn#whole_genome_genotyping_and_copy_number_variation_analysis
• Genotyping & CNV: http://www.illumina.com/applications.ilmn#custom_low_to_mid_plex_genotyping
• Epigenetics: http://www.illumina.com/applications.ilmn#gene_regulation_and_epigenetic_analysis
• Cancer: http://www.illumina.com/applications.ilmn#cancer
• Gene Regulation & Epigenetics: http://www.illumina.com/applications.ilmn#gene_regulation_and_epigenetic_analysis
• Software: http://www.illumina.com/applications/sequencing.ilmn#Software